The first assignment is to create the ordinal logit from scratch. The probability of person going to physician never, once, ..., five times or more than five times is modelled.
The second assignment is a selection model. It uses the semiparametric version and in the first step the nonparametric probit is used for logical consistency, i.e. if I assume residuals are not normally distributed and cannot use standard parametric regression in second step I shouldn't in the first either. Naturally the last step was bootstrapping for standard errors. Influence of insurance on health expenditures is studied.